Abstract
Background
Nursing resources, such as staffing ratios and skill mix, vary across hospitals. Better nursing resources have been linked to better patient outcomes but are assumed to increase costs. The value of investments in nursing resources, in terms of clinical benefits relative to costs, is unclear.
Objective
To determine whether there are differential clinical outcomes, costs, and value among medical patients at hospitals characterized by better or worse nursing resources.
Design
Matched cohort study of patients in 306 acute care hospitals.
Patients
A total of 74,045 matched pairs of fee-for-service Medicare beneficiaries admitted for common medical conditions (25,446 sepsis pairs; 16,332 congestive heart failure pairs; 12,811 pneumonia pairs; 10,598 stroke pairs; 8858 acute myocardial infarction pairs). Patients were also matched on hospital size, technology, and teaching status.
Main Measures
Better (n = 76) and worse (n = 230) nursing resourced hospitals were defined by patient-to-nurse ratios, skill mix, proportions of bachelors-degree nurses, and nurse work environments. Outcomes included 30-day mortality, readmission, and resource utilization-based costs.
Key Results
Patients in hospitals with better nursing resources had significantly lower 30-day mortality (16.1% vs 17.1%, p < 0.0001) and fewer readmissions (32.3% vs 33.6%, p < 0.0001) yet costs were not significantly different ($18,848 vs 18,671, p = 0.133). The greatest outcomes and cost advantage of better nursing resourced hospitals were in patients with sepsis who had lower mortality (25.3% vs 27.6%, p < 0.0001). Overall, patients with the highest risk of mortality on admission experienced the greatest reductions in mortality and readmission from better nursing at no difference in cost.
Conclusions
Medicare beneficiaries with common medical conditions admitted to hospitals with better nursing resources experienced more favorable outcomes at almost no difference in cost.
Electronic supplementary material
The online version of this article (10.1007/s11606-020-06151-z) contains supplementary material, which is available to authorized users.
KEY WORDS: health services research, nursing, quality of care
Introduction
Hospitals are under pressure to improve patient care quality amidst thinning financial margins. Clinicians are committed to the best possible outcomes for their patients. Administrators, however, must weigh the cost-quality tradeoffs of investments. A recent study found good value among major teaching hospitals by demonstrating lower mortality at slightly higher cost for medical patients in teaching hospitals as compared with similar patients in non-teaching hospitals.1 Only 13% of US hospitals are major teaching hospitals with medical residents.1 In contrast, every hospital employs nurses, yet we know little about the comparative value of investments in nursing resources, which are undertaken to different extents by all hospitals. In this study, we examine hospitals that have made substantial versus limited investments in nursing, comparing the costs of those investments relative to the outcomes for medical patients.
Improvements in nursing resources may pay for themselves through better care quality. For example, increasing nurse staffing has been associated with fewer deaths, adverse outcomes, nosocomial infections, and shorter lengths of stay—with associated cost savings.2–4 More favorable nursing environments have been associated with greater value among surgical cohorts, especially for the sickest patients.5, 6 Increasing the proportion of nurses educated at the baccalaureate level is linked to shorter lengths of stay, fewer readmissions, and consequently lower costs.7, 8
In this study, we compare clinically and demographically similar patients admitted to comparable hospitals with respect to size, teaching status, and technology capabilities, but that are otherwise dissimilar with respect to their nursing resources. We match pairs of patients admitted for the same medical condition (i.e., sepsis, congestive heart failure, pneumonia, stroke, acute myocardial infarction), and then closely match the pairs to be as similar as possible on over 60 comorbidities and demographic characteristics. Patient pairs are exactly matched on categories of hospital size, teaching, and technology status. Thus, we were able to evaluate whether clinical outcomes and costs of care differ in function of nursing resources.
Our findings inform the economic case for hospital nursing by evaluating whether the clinical outcomes and costs of care differed across two distinct types of hospitals—those with better versus worse nursing resources defined by patient-to-nurse ratios, proportion of bachelor’s-educated nurses, nursing skill mix, and nurse work environment. We make apparent the costs and associated clinical benefits of hospital nursing resources for patients with common medical conditions and varying degrees of clinical severity. These findings are important to clinicians who seek the best clinical outcomes for their patients and can be used to inform resource allocation decisions made by hospital administrators as well as policy decisions that impact patient care.
METHODS
Patient Sample
The patient sample includes Medicare fee-for-service beneficiaries, 65.5 years or older, who were admitted to an acute care hospital in California, Florida, New Jersey, or Pennsylvania between January 1, 2013, and September 30, 2015, for one of five medical conditions: acute myocardial infarction, congestive heart failure, pneumonia, stroke, or sepsis. Patients missing age, sex, valid date of death, or who were enrolled in an HMO or who lacked Part B coverage in the 6 months prior to admission, are excluded. Patient data includes the following CMS Research Identifiable Files: inpatient, outpatient, carrier (physician Part B), hospice, skilled nursing facility, durable medical equipment, and the master beneficiary summary file.
The index admission is defined as the first medical admission during the study period to a study hospital. A 180-day look-back across all utilization files was completed to identify patient characteristics including age, sex, race, transfer-in status, emergent admission, and 64 comorbidities (Appendix 1). Using patient covariates, we constructed propensity scores for patients’ likelihood of receiving care in a hospital with better nursing resources. With a 10% random sample of patient data external to our matched sample, we generated a 30-day mortality risk model to determine study patients’ probability of 30-day mortality (Appendix 2).9, 10
Hospital Characteristics
Hospitals with better and worse nursing resources are defined using four dimensions of nursing: patient-to-nurse staffing, skill mix, education, and the nurse work environment. These measures were constructed from the 2016 RN4CAST-US study, a large survey of registered nurses (RNs) in 4 states (i.e., CA, FL, NJ, PA) who reported on organizational features and resources of their hospitals.11
Nurse responses pertain to the last shift the nurse worked and were aggregated within hospitals to create hospital measures of nursing. Patient-to-nurse staffing is the number of patients per direct care RN on medical-surgical (or equivalent) units. RNs were asked to report the total number of RNs, licensed practice nurses (LPNs), and unlicensed assistive personnel (UAPs), on their unit during their last shift. Skill mix was calculated as the proportion of RNs to all nursing personnel (i.e., RNs, LPNs, UAPs). Education is the proportion of RNs with at least a bachelor’s degree in nursing. The nurse work environment is measured using the 31-item Practice Environment Scale of the Nursing Work Index (PES-NWI)—a National Quality Forum endorsed scale.12 The subscales of the PES-NWI measure aspects of the nurse environment, including nurse-physician relations, staffing and resource adequacy, and nursing leadership.
Study hospitals were categorized by their size and teaching status, using Healthcare Cost Report Information System (HCRIS) data. Small hospitals had < 250 beds; large hospitals had ≥ 250 beds. Teaching status was categorized as non-teaching (< 0.05 residents to bed, RB), minor (≥ 0.05 and < 0.25 RB), and major (≥ 0.25 RB). Hospitals were also categorized by their technology capabilities, using the Medicare Provider of Service file. High technology hospitals had the capacity to perform major organ transplant and/or open-heart surgery.
Outcomes
Patient outcome measures were 30-day mortality, 30-day readmission (or death), length of stay, intensive care unit (ICU) admission, and ICU length of stay. Economic performance was evaluated using a measure of 30-day resource utilization-based costs, which were inflation-adjusted and computed from the costs of resources used during the in-hospital admission as well as any costs accrued 30-days post-admission (Appendix 4).5, 6, 13, 14 If a patient was readmitted within 30 days of the index admission date, all costs accrued during the readmission (including in-patient costs beyond 30 days) were included. In-hospital costs are a function of length of stay, level of care (i.e., ICU vs. general unit),15 and total relative value units from bills.16 Our resource utilization-based measure of cost uses standardized national prices for resources to allow for meaningful comparisons between hospitals, which has advantages over alternative cost measures, such as cost-to-charge ratios based on negotiated pricing agreements.
Thirty-day costs were adjusted to reflect the costs of higher levels of nurse staffing and skill mix in better nursing resourced hospitals (i.e., 30-day nurse-adjusted costs). For each hospital, costs were adjusted to reflect whether the hospital was above or below the average staffing and to account for salary differences based on skill mix composition. The adjustments were calculated using national average nurse salary data from the Bureau of Labor and Statistics and included adjustments for benefits (Appendix 4). Based on the hospital’s nursing costs per day, patient costs were concomitantly adjusted.
Statistical Analysis
Defining Better and Worse Nursing Resourced Hospitals
We defined study hospitals as having either better or worse nursing resources using a coherence rank score based on four aspects of nursing (i.e., staffing, skill mix, education, work environment) (Appendix 5).6, 17 The four aspects of nursing within each of the 512 study hospitals were compared to all other hospitals to compute a score for every hospital using coherence methods.18 In brief, each hospital was assigned a score between 1 and 512 and rank-ordered. Hospitals in the top 15% were defined as the better nursing resourced hospitals. Hospitals in the bottom 45% were defined as the worse nursing resource hospitals.
Matching Algorithm
Using DesignMatch in R, we built an algorithm to match each patient in a better nursing resourced hospital with a clinically and demographically similar patient in a worse nursing resourced hospital.19–21 To ensure the matched patient pairs were as similar as possible on measurable clinical and demographic variables, we required patients to be exactly matched on ICD-9 principal diagnosis codes (Appendix 3), 30-day mortality risk score quintile, and hospital characteristics (i.e., categories of size, teaching status, technology capability). After exact matching on these variables, we used fine balance22 and distance minimization techniques to create patient pairs that are as similar as possible on over 60 covariates. These included age, sex, race, emergent admission status, transfer-in status, propensity score for attending a better nursing resourced hospital, 64 comorbidities, and continuous measures of 30-day mortality risk score, hospital size, and resident-to-bed ratio (Appendix 1).
Quality of the patient matches was assessed using standardized differences (SD) with a goal of less than 0.1 SD on all covariates.23 Binary outcomes were compared within patient pairs using the McNemar test.24 Continuous outcomes were reported using m-statistics similar to a 1% trim for each tail.25, 26
RESULTS
Characteristics of Better and Worse Nursing Resourced Hospitals
Prior to matching patients, we defined hospitals as having better or worse nursing resources based on four dimensions of nursing. Each of these four dimensions were superior in the better nursing resourced hospitals as compared with the worse nursing resourced hospitals (Table 1). On average, nurses in better nursing resourced hospitals cared for 1.5 fewer patients at a time (4.30 vs 5.79, p < 0.0001), had a richer skill mix of RNs to total nursing staff (0.85 vs 0.78, p < 0.0001), had a greater proportion of nurses with at least a bachelor’s degree (67.5% vs 43.2%, p < 0.0001), and reported more favorable work environments (3.01 vs 2.68, p < 0.0001), as compared to the worse nursing resourced hospitals.
Table 1.
Differences in Nursing Resources
| Characteristic | Better nursing resourced hospitals | Worse nursingresourced hospitals | p value |
|---|---|---|---|
| N = 76 | N = 230 | ||
| Patient-to-nurse staffing ratio, mean (SD) | 4.30 (0.50) | 5.79 (1.05) | < 0.0001 |
| Nurse skill mix, mean (SD) | 0.85 (0.04) | 0.78 (0.05) | < 0.0001 |
| Proportion of nurses with a BSN or higher, % | 67.45% | 43.24% | < 0.0001 |
| Nurse work environment, mean (SD) | 3.01 (0.19) | 2.68 (0.22) | < 0.0001 |
Bachelors of Science in Nursing (BSN)
Quality of Patient Matches
Prior to matching patients, our initial sample consisted of 86,609 medical patients in the 76 better nursing resourced hospitals and 276,857 medical patients in the 230 worse nursing resourced hospitals. After matching patients exactly on principal diagnosis code, mortality risk score quintile, hospital size category, teaching status category, and technology category, and then balancing on over 60 covariates, our analytic patient sample consisted of 74,045 matched patient pairs. Table 2 presents selected covariates, before and after matching patients. Standardized differences between patient pairs did not exceed 0.1 (Appendix 6).
Table 2.
Selected Matched Patient and Hospital Characteristics
| Characteristics (percent unless noted) | Before match | Matched | Matched | Before match | Before match | After match |
|---|---|---|---|---|---|---|
| Patients in better nursing resourced hospitals | Patients in better nursing resourced hospitals | Patients in worse nursing resourced hospitals | Patients in worse nursing resourced hospitals | Standardized difference | Standardized difference | |
| (n = 86,609) | (n = 74,045) | (n = 74,045) | (n = 276,857) | |||
| Patient characteristics | ||||||
| Age, mean (years) | 80.3 | 80.6 | 80.7 | 80.7 | − 0.05 | − 0.01 |
| Male | 47.6 | 47.1 | 46.9 | 47.7 | 0.00 | 0.00 |
| Black | 8.2 | 6.9 | 7.3 | 5.3 | 0.11 | − 0.01 |
| Hispanic | 13.3 | 14.3 | 13.4 | 6.3 | 0.24 | 0.03 |
| Probability of 30-day mortality | 17.5 | 17.3 | 17.2 | 16.0 | 0.11 | 0.01 |
| Emergency admission | 76.9 | 74.9 | 75.8 | 85.4 | − 0.22 | − 0.02 |
| Transfer-in | 0.6 | 0.6 | 0.7 | 0.7 | − 0.02 | − 0.01 |
| Comorbidities | ||||||
| Hypertension | 84.7 | 85.1 | 87.2 | 87.5 | − 0.08 | − 0.06 |
| Diabetes | 44.3 | 44.2 | 46.4 | 45.0 | − 0.01 | − 0.04 |
| Renal failure | 43.7 | 43.6 | 43.6 | 41.5 | 0.04 | 0.00 |
| Congestive heart failure | 38.8 | 38.6 | 37.7 | 38.5 | 0.00 | 0.02 |
| Chronic obstructive pulmonary disease | 32.9 | 33.3 | 35.0 | 38.7 | − 0.12 | − 0.04 |
| Stroke | 12.3 | 12.2 | 11.4 | 11.5 | 0.02 | 0.02 |
| Past AMI | 3.8 | 3.8 | 3.7 | 4.0 | − 0.01 | 0.00 |
| Medical conditions | ||||||
| Acute myocardial infarction | 11.4 | 12.0 | 12.0 | 14.2 | − 0.08 | 0.00 |
| Congestive heart failure | 22.1 | 22.1 | 22.1 | 24.0 | − 0.05 | 0.00 |
| Pneumonia | 16.8 | 17.3 | 17.3 | 20.5 | − 0.09 | 0.00 |
| Stroke | 35.6 | 34.4 | 34.4 | 27.9 | 0.16 | 0.00 |
| Sepsis | 14.1 | 14.3 | 14.3 | 13.4 | 0.02 | 0.00 |
| Hospital characteristics | ||||||
| Small (< 250 beds) | 25.7 | 30.0 | 30.0 | 47.6 | − 0.47 | 0.00 |
| Large (≥ 250 beds) | 74.3 | 70.0 | 70.0 | 52.4 | 0.47 | 0.00 |
| Beds, mean | 422.5 | 383.0 | 375.3 | 295.0 | 0.63 | 0.04 |
| Non-teaching | 50.1 | 58.6 | 58.6 | 65.0 | − 0.30 | 0.00 |
| Minor teaching | 21.2 | 24.8 | 24.8 | 28.3 | − 0.17 | 0.00 |
| Major teaching | 28.7 | 16.7 | 16.7 | 6.7 | 0.60 | 0.00 |
| Resident to bed ratio, mean | 0.23 | 0.13 | 0.12 | 0.07 | 0.61 | 0.05 |
| High technology status | 84.2 | 81.5 | 81.5 | 49.0 | 0.80 | 0.00 |
| Nursing resources | ||||||
| Patient-to-nurse staffing ratio, mean | 4.37 | 4.40 | 5.52 | 5.77 | − 1.93 | − 1.55 |
| Nurse skill mix, mean | 0.84 | 0.85 | 0.77 | 0.78 | 1.55 | 1.69 |
| Proportion of nurses with a BSN or higher | 68.70 | 67.21 | 47.35 | 44.77 | 2.00 | 1.66 |
| Nurse work environment, mean | 3.04 | 3.03 | 2.73 | 2.70 | 1.67 | 1.49 |
After matching, the number of patients in the better resourced hospitals declines from 86,609 to 74,045 because not all patients in the best resourced hospitals could be matched with patients in a worse resourced hospital. The standardized differences after matching among the four measures of nursing resources are large (as expected) since the matching algorithm selected patient pairs who were in distinctly different hospitals with respect to their nursing resources
Prior to matching, the patients in better nursing resourced hospitals were more likely to be in large, high-technology, major teaching hospitals. Our matching algorithm required patients to be matched exactly on size, teaching, and technology categories, as reflected by the standardized difference after matching of 0.00. Further balance within size and teaching status categories was achieved through distance minimization techniques on the continuous measures of hospital’s number of beds and RB ratios (Table 2).
Outcomes
Table 3 presents a comparison of outcomes between patients in better and worse nursing resourced hospitals. Among all patients (i.e., the aggregate of all medical conditions), the odds of 30-day mortality and 30-day readmission were each 7% lower for patients in the better nursing resourced hospitals as compared to matched patients in the worse nursing resourced hospitals. The likelihood of being admitted to an ICU was 6% higher in a better nursing hospital and the average length of stay in the ICU was slightly longer although not clinically meaningful (0.99 days in better nursing resourced hospitals versus 0.97 days in worse nursing resourced hospitals, p = 0.031). The overall average length of stay was shorter for patients in better nursing resourced hospitals (5.38 days vs 5.66 days, p < 0.001). Pairwise differences in 30-day resource utilization-based costs were lower in hospitals with better versus worse nursing resources ($18,436 vs $18,708, p = 0.004). However, after accounting for the higher costs of better nurse staffing and greater nursing skill mix, differences in the 30-day nurse-adjusted costs for patients cared for in either the better or worse nursing resourced hospitals were insignificant ($18,848 vs $18,671, p = 0.133).
Table 3.
Comparison of Outcomes Between Matched Patients in Better and Worse Nursing Resourced Hospitals
| Outcome | Better nursing resourced hospitals | Worse nursing resourced hospitals | ||
|---|---|---|---|---|
| All conditions | ||||
| Number of patients | n = 74,045 | n = 74,045 | ||
| % | % | Odds ratio (95% CI) | p value | |
| 30-day mortality | 16.14 | 17.06 | 0.93 (0.90 to 0.96) | < .0001 |
| 30-day readmission (or death) | 32.26 | 33.63 | 0.93 (0.91 to 0.96) | < .0001 |
| ICU admission | 26.92 | 25.97 | 1.06 (1.03 to 1.09) | < .0001 |
| m-estimate | m-estimate | Paired difference (95% CI) | p value | |
| Length of stay (days) | 5.38 | 5.66 | − 0.27 (− 0.32 to − 0.23) | < 0.001 |
| ICU length of stay (days) | 0.99 | 0.97 | 0.03 (0.00 to 0.05) | 0.031 |
| 30-day costs ($) | 18,436 | 18,708 | − 285 (− 481 to − 88) | 0.004 |
| 30-day nurse-adjusted costs ($) | 18,848 | 18,671 | 153 (− 46 to 351) | 0.133 |
| Sepsis | ||||
| Number of patients | n = 25,446 | n = 25,446 | ||
| % | % | Odds ratio (95% CI) | p value | |
| 30-day mortality | 25.31 | 27.63 | 0.88 (0.84 to 0.91) | < .0001 |
| 30-day readmission (or death) | 41.15 | 43.79 | 0.89 (0.85 to 0.92) | < .0001 |
| ICU admission | 36.93 | 37.82 | 0.96 (0.93 to 1.00) | 0.034 |
| m-estimate | m-estimate | Paired difference (95% CI) | p value | |
| Length of stay (days) | 6.66 | 6.98 | − 0.33 (− 0.42 to − 0.23) | < 0.001 |
| ICU length of stay (days) | 1.63 | 1.70 | − 0.07 (− 0.13 to − 0.01) | 0.028 |
| 30-day costs ($) | 22,744 | 23,547 | − 819 (− 1237 to − 399) | < 0.001 |
| 30-day nurse-adjusted costs ($) | 23,238 | 23,680 | − 459 (− 884 to − 35) | 0.034 |
| Congestive heart failure | ||||
| Number of patients | n = 16,332 | n = 16,332 | ||
| % | % | Odds ratio (95% CI) | p value | |
| 30-day mortality | 10.03 | 10.00 | 1.00 (0.93 to 1.08) | 0.925 |
| 30-day readmission (or death) | 29.88 | 30.33 | 0.98 (0.93 to 1.03) | 0.370 |
| ICU admission | 14.90 | 12.23 | 1.26 (1.18 to 1.35) | < .0001 |
| m-estimate | m-estimate | Paired difference (95% CI) | p value | |
| Length of stay (days) | 4.90 | 5.14 | − 0.25 (− 0.33 to − 0.16) | < 0.001 |
| ICU length of stay (days) | 0.48 | 0.36 | 0.12 (0.09 to 0.16) | < 0.001 |
| 30-day costs ($) | 15,979 | 15,714 | 187 (− 173 to 547) | 0.308 |
| 30-day nurse-adjusted costs ($) | 16,391 | 15,548 | 773 (408 to 1,137) | < 0.001 |
| Pneumonia | ||||
| Number of patients | n = 12,811 | n = 12,811 | ||
| % | % | Odds ratio (95% CI) | p value | |
| 30-day mortality | 10.32 | 10.50 | 0.98 (0.90 to 1.06) | 0.628 |
| 30-day readmission (or death) | 23.92 | 24.61 | 0.96 (0.91 to 1.02) | 0.179 |
| ICU admission | 10.25 | 9.71 | 1.06 (0.98 to 1.16) | 0.145 |
| m-estimate | m-estimate | Paired difference (95% CI) | p value | |
| Length of stay (days) | 4.88 | 5.25 | − 0.37 (− 0.46 to − 0.29) | < 0.001 |
| ICU length of stay (days) | 0.34 | 0.35 | 0.00 (− 0.03 to 0.04) | 0.892 |
| 30-day costs ($) | 13,664 | 14,349 | − 703 (− 1034 to − 372) | < 0.001 |
| 30-day nurse-adjusted costs ($) | 13,992 | 14,205 | − 220 (− 554 to 114) | 0.197 |
| Stroke | ||||
| Number of patients | n = 10,598 | n = 10,598 | ||
| % | % | Odds ratio (95% CI) | p value | |
| 30-day mortality | 13.38 | 14.18 | 0.93 (0.85 to 1.01) | 0.075 |
| 30-day readmission (or death) | 25.90 | 26.84 | 0.95 (0.89 to 1.01) | 0.108 |
| ICU admission | 26.32 | 23.53 | 1.17 (1.10 to 1.25) | <.0001 |
| m-estimate | m-estimate | Paired difference (95% CI) | p value | |
| Length of stay (days) | 4.28 | 4.47 | − 0.18 (− 0.27 to − 0.09) | < 0.001 |
| ICU length of stay (days) | 0.71 | 0.71 | 0.02 (− 0.03 to 0.07) | 0.405 |
| 30-day costs ($) | 15,455 | 15,463 | 10 (− 378 to 398) | 0.960 |
| 30-day nurse-adjusted costs ($) | 15,819 | 15,434 | 392 (2 to 784) | 0.049 |
| Acute myocardial infarction | ||||
| Number of patients | n = 8,858 | n = 8,858 | ||
| % | % | Odds ratio (95% CI) | p value | |
| 30-day mortality | 12.78 | 12.60 | 1.02 (0.93 to 1.12) | 0.706 |
| 30-day readmission (or death) | 30.74 | 31.67 | 0.95 (0.89 to 1.02) | 0.167 |
| ICU admission | 45.16 | 43.69 | 1.07 (1.01 to 1.14) | 0.034 |
| m-estimate | m-estimate | Paired difference (95% CI) | p value | |
| Length of stay (days) | 4.75 | 4.87 | − 0.14 (− 0.26 to − 0.01) | 0.030 |
| ICU length of stay (days) | 1.46 | 1.29 | 0.16 (0.08 to 0.24) | < 0.001 |
| 30-day costs ($) | 21,368 | 20,606 | 675 (33 to 1,315) | 0.039 |
| 30-day nurse-adjusted costs ($) | 21,705 | 20,530 | 1,090 (442 to 1,736) | < 0.001 |
p values for outcomes were calculated using McNemar test for binary outcomes and m-statistics for continuous ones
Clinical outcomes and cost findings differed by medical condition. For example, significantly lower 30-day mortality and 30-day readmissions were observed in better nursing resourced hospitals for patients with sepsis. Shorter lengths of stay were observed among all conditions: sepsis, stroke, pneumonia, congestive heart failure, and acute myocardial infarction. ICU admission was significantly lower for sepsis patients in better nursing resourced hospitals, but higher among patients with congestive heart failure, stroke, and acute myocardial infarction. Nurse-adjusted costs were significantly lower in the better nursing resourced hospitals for patients with sepsis ($23,238 vs $23,680, p = 0.034), and not significantly different for patients with pneumonia. The pairwise difference in cost was marginally higher in the better nursing hospitals for patients with congestive heart failure ($16,391 vs $15,548, p < 0.001), stroke ($15,819 vs $15,434, p < 0.049), and acute myocardial infarction ($21,705 vs $20,530, p < 0.001), perhaps because the rates of admission to the ICU were also higher for patients with these conditions in the better nursing resourced hospitals. Results for in-hospital costs are reported in Appendix 7.
Outcomes by Patient Clinical Risk
To determine whether there were differences in outcomes and costs by patient clinical risk on admission, we aggregated the medical conditions (Table 4). Overall, patients in better nursing resourced hospitals had better clinical outcomes as compared to patients in worse nursing resourced hospitals, and differences were greatest among patients with higher clinical risk. For example, the mortality difference was 1.99 percentage points lower for the highest risk patients in better as compared to worse nursing resourced hospitals, while there were no significant mortality differences observed among the lowest risk patients. ICU admission rates were not significantly different for patients in the 4th and 5th risk quintiles. Instead, the greatest differences were observed among patients in the lowest quintile. Nurse-adjusted costs were no different between patients in the better and worse nursing resourced hospitals, except for the lowest risk patients, who had higher costs in the better nursing resourced hospitals as compared to the worst ($14,638 vs $14,007, p < 0.001). The difference of $607 is marginal, at less than 5% of the overall cost. The difference in cost is likely associated with greater ICU utilization among low risk patients in better nursing resourced hospitals.
Table 4.
Comparison of Outcomes Between Matched Cases in Better and Worse Nursing Hospitals Across Risk Quintiles
| Patient risk quintiles of 30-day mortality | ||||||
|---|---|---|---|---|---|---|
| Outcome | Overall | Q1 (lowest) | Q2 | Q3 | Q4 | Q5 (highest) |
| (n = 74,045) | (n = 13,561) | (n = 14,156) | (n = 14,440) | (n = 15,433) | (n = 16,455) | |
| 30-day mortality, % | ||||||
| Better nursing | 16.14 | 3.19 | 6.78 | 11.33 | 19.67 | 35.77 |
| Worse nursing | 17.06 | 3.22 | 6.94 | 12.18 | 20.98 | 37.76 |
| Difference | − 0.92*** | − 0.02 | − 0.16 | − 0.85* | − 1.31** | − 1.99*** |
| 30-day readmission (or death), % | ||||||
| Better nursing | 32.26 | 17.65 | 22.81 | 27.55 | 36.53 | 52.55 |
| Worse nursing | 33.63 | 18.29 | 23.83 | 28.86 | 38.61 | 54.20 |
| Difference | − 1.37*** | − 0.64 | − 1.02* | − 1.32* | − 2.08*** | − 1.65** |
| ICU admission, % | ||||||
| Better nursing | 26.92 | 18.97 | 22.13 | 25.37 | 29.98 | 36.07 |
| Worse nursing | 25.97 | 16.09 | 20.44 | 24.27 | 29.62 | 36.93 |
| Difference | 0.95*** | 2.88*** | 1.69*** | 1.11* | 0.36 | − 0.86 |
| Length of stay, d (m-estimate) | ||||||
| Better nursing | 5.38 | 4.10 | 4.66 | 5.35 | 5.94 | 6.63 |
| Worse nursing | 5.66 | 4.30 | 5.02 | 5.62 | 6.25 | 6.82 |
| Difference | − 0.27*** | − 0.20*** | − 0.36*** | − 0.28*** | − 0.32*** | − 0.21*** |
| ICU length of stay, d (m-estimate) | ||||||
| Better nursing | 0.99 | 0.51 | 0.68 | 0.94 | 1.21 | 1.56 |
| Worse nursing | 0.97 | 0.43 | 0.63 | 0.88 | 1.19 | 1.59 |
| Difference | 0.03* | 0.09*** | 0.06** | 0.04 | 0.01 | − 0.05 |
| 30-day cost, $ (m-estimate) | ||||||
| Better nursing | 18,436 | 14,304 | 16,085 | 18,202 | 20,173 | 22,513 |
| Worse nursing | 18,708 | 14,119 | 16,362 | 18,381 | 20,703 | 22,947 |
| Difference | − 285** | 182 | − 343 | − 206 | − 575* | − 453 |
| 30-day nurse-adjusted cost, $ (m-estimate) | ||||||
| Better nursing | 18,848 | 14,638 | 16,443 | 18,628 | 20,587 | 23,039 |
| Worse nursing | 18,671 | 14,007 | 16,270 | 18,371 | 20,723 | 22,997 |
| Difference | 153 | 607*** | 89 | 212 | − 153 | 41 |
p values: < 0.05*, < 0.01**, < 0.001***. Numbers are rounded for display. For binary outcomes, the difference is the difference in rates. For continuous outcomes, the difference is the m-estimate of the typical pair difference. Being nonlinear, the m-estimate of the difference is not the difference of the m-estimates
Value
We define value as the pairwise difference in nurse-adjusted cost compared to the difference in 30-day mortality. Among the medical conditions overall, the difference in nurse-adjusted costs was small with an estimate of $153 and confidence interval of − $46 to $351, and mortality rates were more favorable in the better nursing resourced hospitals; thus, there is a clear value case for better nursing resources. Among patients with sepsis, lower mortality was achieved in better nursing resourced hospitals at lower costs, even after adjusting for the additional costs of better nursing resources. Among the other patients, there were no observed significant differences in mortality and thus, a value estimate could not be defined.
DISCUSSION
Hospitals with superior nursing resources, including better patient-to-nurse staffing ratios, a greater skill mix of RNs to all nursing staff, a greater proportion of bachelor’s-degree nurses, and more favorable nurse work environments, have better patient outcomes at no difference in costs (pairwise difference $153, 95% CI − $46 to $351, p = 0.133) even after accounting for the higher costs associated with better nursing resources.
This study demonstrates that better clinical outcomes, including lower mortality, are achieved at no significant difference in costs among medical Medicare patients in hospitals that invest in their nursing human capital. An analysis of surgical Medicare patients similarly found lower mortality, failure-to-rescue, shorter lengths of stay and marginally higher costs among patients in better nursing resourced hospitals.6 Other studies of medical patients have linked nursing resources with better outcomes including lower mortality,27 lower readmissions,28 and fewer complications.29, 30 A study examining costs of care among medical patients in Veteran’s Administration facilities and its association with nurse hours per patient day and skill mix found that greater time RNs spend with patients is associated with cost offsets derived from shorter lengths of stay.31 In our study, medical patients in better nursing resourced hospitals also had shorter lengths of stay as compared with their matched pairs in the worse resourced hospitals.
We found the greatest differences in clinical outcomes and costs among patients admitted for sepsis. Initiatives like the Surviving Sepsis Campaign have been a positive step towards quality improvement, but despite widespread understanding of how to care for septic patients, significant variation in outcomes remains. Treatment and management of sepsis uses standardized management protocols,32 which are primarily driven by the bedside nurse responsible for surveilling patients’ response to treatment, titrating medications, and clinical decision-making. Variation in sepsis mortality outcomes across hospitals has been attributed to the hospital where the patient received care.33–35 Our finding that sepsis outcomes and costs of care were superior in hospitals with better nursing resources suggests that attention to system-level attributes of nursing is important to improving sepsis outcomes. Whether through lower mortality, fewer readmissions, or shorter lengths of stay, patients in each of the medical condition categories in our study experienced a benefit from superior nursing resources.
Limitations
Our findings should be considered in the context of its limitations. The patient sample included only Medicare beneficiaries with one of five common medical conditions and therefore does not represent the full range of patients cared for in hospitals. Our measurements of nursing resources are cross-sectional, which limits our ability to make causal inferences, although panel studies have shown that improving nurse work environments and increasing the percentage of bachelors-educated nurses over time are associated with improvements in patient outcomes.36–39 Although we matched on over 60 patient and hospital characteristics, we were only able to match on observable characteristics using Medicare administrative data, and thus, our results are potentially subject to unmeasured confounding. In Appendix 8, we assess sensitivity to unmeasured confounding and show that it is unlikely an explanation for our results. Finally, we use standardized national pricing. Although others have used cost-to-charge ratios, they have found similar results.2–4
CONCLUSION
Patients admitted for common medical conditions in hospitals with better nursing resources had a lower likelihood of mortality, readmission, and shorter lengths of stay, with almost no difference in costs as compared to matched patients in hospitals with worse nursing resources. Our findings suggest that investments in nursing resources are associated with better clinical outcomes and almost no difference in cost.
Electronic supplementary material
(DOCX 209 kb)
Funders
This study was supported by grants from the National Institute of Nursing Research, NIH (R01 NR016019, Silber & McHugh; R01 NR014855, Aiken).
Compliance with Ethical Standards
Conflict of Interest
The authors declare that they do not have a conflict of interest.
Footnotes
Prior presentations
This study was accepted for presentation at the Annual Research Meeting of AcademyHealth on June 16, 2020, but the presentation was cancelled due to COVID-19.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Silber JH, Rosenbaum PR, Niknam B, et al. Comparing outcomes and costs of medical patients treated at major teaching and nonteaching hospitals: a national matched analysis. J Gen Intern Med. 2020;35:743–52. doi: 10.1007/s11606-019-05449-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Dall TM, Chen YJ, Seifert RF, Maddox PJ, Hogan PF. The economic value of professional nursing. Med Care. 2009;47:97–104. doi: 10.1097/MLR.0b013e3181844da8. [DOI] [PubMed] [Google Scholar]
- 3.Needleman J, Buerhaus PI, Stewart M, Zelevinsky K, Mattke S. Nurse staffing in hospitals: Is there a business case for quality? Health Aff (Millwood). 2006;25:204–11. doi: 10.1377/hlthaff.25.1.204. [DOI] [PubMed] [Google Scholar]
- 4.Martsolf GR, Auerbach D, Benevent R, et al. Examining the value of inpatient nurse staffing: An assessment of quality and patient care costs. Med Care. 2014;52:982–8. doi: 10.1097/MLR.0000000000000248. [DOI] [PubMed] [Google Scholar]
- 5.Silber JH, Rosenbaum PR, McHugh MD, et al. Comparison of the value of nursing work environments in hospitals across different levels of patient risk. JAMA Surg. 2016;151:527–36. doi: 10.1001/jamasurg.2015.4908. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lasater KB, McHugh MD, Rosenbaum PR, et al. Valuing hospital investments in nursing: a multi-state matched cohort study of surgical patients. BMJ Qual Saf. 10.1136/bmjqs-2019-010534. [Epub ahead of print]. [DOI] [PMC free article] [PubMed]
- 7.Yakusheva O, Lindrooth R, Weiss M. Economic evaluation of the 80% baccalaureate nurse workforce recommendation: A patient-level analysis. Med Care. 2014;52:864–9. doi: 10.1097/MLR.0000000000000189. [DOI] [PubMed] [Google Scholar]
- 8.Yakusheva O, Lindrooth R, Weiss M. Nurse value-added and patient outcomes in acute care. Health Serv Res. 2014;49:1767–86. doi: 10.1111/1475-6773.12236. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hansen BB. The prognostic analogue of the propensity score. Biometrika. 2008;95:481–88. doi: 10.1093/biomet/asn004. [DOI] [Google Scholar]
- 10.Rosenbaum P, Rubin D. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55. doi: 10.1093/biomet/70.1.41. [DOI] [Google Scholar]
- 11.Lasater KB, Jarrin OF, Aiken LH, McHugh MD, Sloane DM, Smith HL. A methodology for studying organizational performance: A multi-state survey of front line providers. Med Care. 2019;57:742–49. doi: 10.1097/MLR.0000000000001167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lake ET. The nursing practice environment: measurement and evidence. Med Care Res Rev. 2007;64:104S–22S. doi: 10.1177/1077558707299253. [DOI] [PubMed] [Google Scholar]
- 13.Silber JH, Rosenbaum PR, Kelz RR, et al. Medical and financial risks associated with surgery in the elderly obese. Ann Surg. 2012;256:79–86. doi: 10.1097/SLA.0b013e31825375ef. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Bamezai A, Melnick G, Nawathe A. The cost of an emergency department visit and its relationship to emergency department volume. Ann Emerg Med. 2005;45:483–90. doi: 10.1016/j.annemergmed.2004.08.029. [DOI] [PubMed] [Google Scholar]
- 15.Halpern NA, Pastores SM. Critical care medicine in the United States 2000-2005: An analysis of bed numbers, occupancy rates, payer mix, and costs. Crit Care Med. 2010;38:65–71. doi: 10.1097/CCM.0b013e3181b090d0. [DOI] [PubMed] [Google Scholar]
- 16.Centers for Medicare & Medicaid Services. PFS Relative Value Files. Available at: https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/PhysicianFeeSched/PFS-Relative-Value-Files.html. Accessed Mar 7, 2019.
- 17.Silber JH, Lorch SA, Rosenbaum PR, et al. Time to send the preemie home? Additional maturity at discharge and subsequent healthcare costs and outcomes. Health Serv Res. 2009;44:444–63. doi: 10.1111/j.1475-6773.2008.00938.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Rosenbaum PR. Coherence in observational studies. Biometrics. 1994;50:368–74. doi: 10.2307/2533380. [DOI] [PubMed] [Google Scholar]
- 19.Zubizarreta JR. Using mixed integer programming for matching in an observational study of kidney failure after surgery. J Am Stat Assoc. 2012;107:1360–71. doi: 10.1080/01621459.2012.703874. [DOI] [Google Scholar]
- 20.Zubizarreta JR, Kilcioglu C, Vielma JP. Package ‘designmatch’. Matched Samples that are Balanced and Representative by Design Version 0.3.0. Last Modified Date: Published June 18, 2018. Available at: https://cran.r-project.org/web/packages/designmatch/designmatch.pdf. Accessed July 16, 2018.
- 21.R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing,2018. Available at: http://www.R-project.org. Accessed Apr 20, 2018.
- 22.Rosenbaum PR, Ross RN, Silber JH. Minimum distance matched sampling with fine balance in an observational study of treatment for ovarian cancer. J Am Stat Assoc. 2007;102:75–83. doi: 10.1198/016214506000001059. [DOI] [Google Scholar]
- 23.Cochran WG, Rubin DB. Controlling bias in observational studies. A review. Sankhya. Series A. 1973;35:417–46. [Google Scholar]
- 24.Bishop YM, Fienberg SE, Holland PW. Discrete Multivariate Analysis: Theory and Practice. Springer Science & Business Media: New York; 2007. [Google Scholar]
- 25.Rosenbaum PR. Sensitivity analysis for m-estimates, tests, and confidence intervals in matched observational studies. Biometrics. 2007;456-64(R package sensitivitymv and sensitivitymw):63. doi: 10.1111/j.1541-0420.2006.00717.x. [DOI] [PubMed] [Google Scholar]
- 26.Huber PJ. Robust Statistics. Hoboken, NJ: John Wiley & Sons; Vol. 523. 2004.
- 27.Aiken LH, Smith HL, Lake ET. Lower Medicare mortality among a set of hospitals known for good nursing care. Med Care. 1994;32:771–87. doi: 10.1097/00005650-199408000-00002. [DOI] [PubMed] [Google Scholar]
- 28.Brooks-Carthon JM, Lasater KB, Rearden J, Holland S, Sloane DM. Unmet nursing care linked to rehospitalizations among older black AMI patients: a cross-sectional study of US hospitals. Med Care. 2016;54:457–65. doi: 10.1097/MLR.0000000000000519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Cimiotti JP, Aiken LH, Sloane DM, Wu ES. Nurse staffing, burnout, and health care-associated infection. Am J Infect Control. 2012;40:486–90. doi: 10.1016/j.ajic.2012.02.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.McHugh MD, Shang J, Sloane DM, Aiken LH. Risk factors for hospital-acquired 'poor glycemic control': A case-control study. Int J Qual Health Care. 2011;23:44–51. doi: 10.1093/intqhc/mzq067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Li YF, Wong ES, Sales AE, et al. Nurse staffing and patient care costs in acute inpatient nursing units. Med Care. 2011;49:708–15. doi: 10.1097/MLR.0b013e318223a9f1. [DOI] [PubMed] [Google Scholar]
- 32.Dellinger RP, Levy MM, Rhodes A, et al. Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock, 2012. Intensive Care Med. 2013;39:165–228. doi: 10.1007/s00134-012-2769-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wang HE, Donnelly JP, Shapiro NI, Hohmann SF, Levitan EB. Hospital variations in severe sepsis mortality. Am J Med Qual. 2015;30:328–36. doi: 10.1177/1062860614534461. [DOI] [PubMed] [Google Scholar]
- 34.Hatfield KM, Dantes RB, Baggs J, et al. Assessing variability in hospital-level mortality among U.S. Medicare beneficiaries with hospitalizations for severe sepsis and septic shock. Crit Care Med. 2018;46:1753–60. doi: 10.1097/CCM.0000000000003324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Corl K, Levy M, Phillips G, Terry K, Friedrich M, Trivedi AN. Racial and ethnic disparities in care following the New York State Sepsis Initiative. Health Aff (Millwood). 2019;38:1119–26. doi: 10.1377/hlthaff.2018.05381. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Kutney-Lee A, Stimpfel AW, Sloane DM, Cimiotti JP, Quinn LW, Aiken LH. Changes in patient and nurse outcomes associated with magnet hospital recognition. Med Care. 2015;53:550–57. doi: 10.1097/MLR.0000000000000355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Kutney-Lee A, Sloane DM, Aiken LH. An increase in the number of nurses with baccalaureate degrees is linked to lower rates of postsurgery mortality. Health Aff (Millwood). 2013;32:579–86. doi: 10.1377/hlthaff.2012.0504. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Aiken LH, Sloane DM, Barnes H, Cimiotti JP, Jarrin OF. McHugh MD. Nurses' and patients' appraisals show patient safety in hospitals remains a concern. Health Aff (Millwood). 2018;37:1744–51. doi: 10.1377/hlthaff.2018.0711. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Sloane DM, Smith HL, McHugh MD, Aiken LH. Effect of changes in hospital nursing resources on improvements in patient safety and quality of care: A Panel Study. Med Care. 2018;56:1001–08. doi: 10.1097/MLR.0000000000001002. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
(DOCX 209 kb)